NVIDIA used SIGGRAPH 2010 to unveil the newest versions of its high-end workstation Quadro line. As expected, the line-up is based on Fermi, the company's next-generation graphics architecture, and is available in mobile as well as various internal options for workstations or Quadro Plex systems.
The full roster announced Tuesday includes:
Mobile
- Quadro 5000M
- 2GB memory
- 76.8GBps bandwidth
- 320 CUDA cores
Workstation cards
- Quadro 4000
- 2GB memory
- 89.6GBps bandwidth
- 256 CUDA cores
- Quadro 5000
- 2.5GB memory
- 120GBps bandwidth
- 352 CUDA cores
- Quadro 6000
- 6GB memory
- 120GBps bandwidth
- 448 CUDA cores
All the workstation cards feature one dual-link DVI and two DisplayPort outputs. At the über-high end is the new Quadro Plex 7000, with 4 dual-link DVI outs, 896 CUDA cores, 12GB of memory, 144GBps bandwidth and a max FSAA of 128x while driving clusters of synced displays.
Aside from the standard Quadro features, the Fermi-based Quadros have some significant differences from the previous generation. There's now full OpenGL 4.1 and DirectX 11 support. The cards also support Shader model 5.0 and sport high-performance, double-precision floats, and ECC memory.
The last two features are mostly for increased accuracy and fault-proofing of GPU-based simulations. While 3D applications are demanding less Quadro-specific support, the rise of GPU-based renderers and increased use of GPUs in science should make the added memory and lower power requirements of the Quadros appealing to a lot of potential users.
We're still waiting to learn about specific cards and price tags from PNY, NVIDIA’s main Quadro manufacturer. Also notably missing from the list is a Mac-compatible Fermi card and, with the new Mac Pros announced Tuesday, it doesn't look like we'll see a GeForce option anytime soon. We'll be speaking with NVIDIA tomorrow, so we'll see if we can dig up some additional details on pricing, shipping, and Mac OS X compatibility.
